The UX Projects of Chris Westgate
Mock3 Copy + Mock2 Copy Mask Copy Copy Copy Copy Copy.png

Hai Dashboard

Tool for customizing AI to business-specific language

My contribution: I designed the UX and UI while working with the data science team to ensure the experience complimented their data science and taught users how to work within the constraints of a language model. Have you ever been frustrated tha

My contribution: I designed the UX and UI while working with the data science team to ensure the experience complimented their data science and taught users how to work within the constraints of a language model.

Have you ever been frustrated that Siri or Alexa didn’t understand a command you asked? What if you could teach these assistants new words and commands? With this project, we created a tool to do exactly that.

With the success of the Hai business analytics app, we decided to create a behind-the-scenes dashboard so any company could customize the AI’s language understanding to suit their needs.

Wireframes

Wireframes

It gets complicated, but basically users would be creating the building blocks of their own language learning model. These building blocks are called Intents and Entities. You can think of them as commands and data types.

You could train the model to understand any question but it needed to know what the important categories, subjects and adjectives were and where to find the relevant corresponding data to answer. So the users would need to link up all this information.

With all these building blocks to keep track of, we decided to color code the various categories in order to keep things organized.

Control panel designs

Control panel designs

Teaching the users how to build their model would be a big challenge. After all, this is work that people go to years of school to learn.

Once we broke things down though, we could take users through each individual step slowly and build up understanding as they went along.

First, they wrote their main intents, then for each intent, several example sentences and questions, then they highlighted the relevant entities in each sentence, and finally they would link entities to the relevant data in their database.

We included a chat feature so one of our data scientists could help out in case users got stuck.

More projects